Comments on: Cervical cancer prevention in El Salvador: A prospective evaluation of screening and triage strategies incorporating high‐resolution microendoscopy to detect cervical precancer

Abstract

Dear editor, We commend the authors for performing a prospective trial to evaluate high-resolution microendoscopy (HRME), human papillomavirus (HPV) testing using careHPV ([QIAGEN, Gaithersberg, MD], a signalamplification assay that detects 14 different carcinogenic HPV types that can be done in roughly 2-5 hours) and visual inspection with acetic acid (VIA) to detect cervical intraepithelial neoplasia (CIN) 2, CIN 3 or cervical cancer (CIN2+). We have concerns with respect to the study methodology, statistical analysis and the use of Proflavine (a fluorescent acridine compound), which was applied to the cervix prior to HRME. In our study, the gold standard for diagnosis of CIN2+ was a combination of colposcopic-directed biopsy, biopsy of VIA-positive sites, biopsy of HRME-positive sites and random biopsy of HRME-negative sites. We reported that, when a five-biopsy gold standard (which includes colposcopic-directed biopsy, random biopsy in cervical quadrants without visible lesions and endocervical curettage [ECC]) was employed, the sensitivity of colposcopic-directed biopsy was 55.5%. When a five-biopsy gold standard was used, the sensitivity for CIN 2+ of VIA (45.9%) was 20% lower than when the gold standard was colposcopic-directed biopsy (65.9%, P < .001). The reason for this 20% difference in sensitivity of VIA is that when a screening test and the gold standard for diagnosing the endpoint are correlated, there is a mathematical equation showing that there is inflation of the sensitivity of the screening test. It would be surprising if HRME were not correlated with colposcopic-directed biopsy (because the site of the HRME reading was defined by abnormal Proflavine or Lugol s iodine staining, positive VIA or positive colposcopic impression). It is unclear how much the additional biopsies of VIA or HRME-positive lesions affected the gold standard of colposcopic-directed biopsy but whatever effect they had may have been overshadowed by the verification bias that they introduced. The statistical problem in this article is that there are only 28 cases of CIN2+; the study power to detect differences in sensitivity of the triage strategies is only 9.6% (α = 0.05). As shown in Table 1 (sensitivity and specificity from Parra s article, 95% CI from an online calculator from University of California San Francisco Clinical and Translational Science Institute), the 95% CI for sensitivity for CIN2+ are so large that one cannot reliably conclude that any of the sensitivities for CIN2+ of the screening algorithms differ (or are similar). Making conclusions concerning the best screening algorithm from this data set is inappropriate. The third problem with our study concerns whether Proflavine, the dye that was applied prior to HRME, is carcinogenic. Proflavine intercalates into DNA and is mutagenic. Whether Proflavine is carcinogenic is unclear. A recent case-control study found the rate of progression to CIN2+ of women that were exposed to Proflavine at colposcopy (0.0%, 0/55) did not differ from those who were not exposed to Proflavine (2.8%, 1/36). But, with median follow-up of 26.3 months, the risk of progression from negative or HPV to CIN3 or cancer (CIN3+) is only 1.9% (47/2490 [95% CI 1.4-2.5]); the power of this case-control study to detect the effect of Proflavine is small. In contrast, Obstoy found that acriflavine (another closely related acridine compound, which intercalates DNA and allows cell nuclei to be imaged with probe-based confocal endomicroscopy) induced considerable DNA damage and concluded that it should not be used in humans, particularly in patients already exposed to respiratory carcinogens. Comparison of VIA, HRME and colposcopic impression is difficult. Although onerous, one way to compare VIA, HRME and colposcopy would be to assign results of colposcopic impression, VIA and HRME